Typhoon Bavi shuts Ningbo port 10–13 July; Shanghai backlog builds

The Analysis

Put the map on the table and start there. Two deep-water gateways sit on either side of Hangzhou Bay. One is fed by the Yangtze and by the whole manufacturing belt standing behind it. The other was built out at sea, on islands that were fishing ground within living memory, tied back to the mainland by one very long bridge. Between 10 and 13 July 2026 both of them stopped. Tropical Cyclone Bavi did that. Ten ports across China, Japan and Taiwan halted operations inside the same four-day window, and six more across China and Taiwan ran disrupted without a full shutdown. Everstream counted them in its 23 July port congestion report, with Hapag-Lloyd and Kuehne+Nagel supplying the operational detail.

Four days is not a long time. I have watched this coast absorb worse than four days. What makes this one worth your attention is not the wind. It is the queue that was still standing ten days later. Kuehne+Nagel's wait-time table in that same 23 July report has Shanghai at 3.2 days of average wait and Ningbo at 2 days. Busan and Yokohama are both at 1.5 days, and Hapag-Lloyd flags vessel bunching and concentrated arrivals at each of them. Ten days after the gates reopened, the backlog had not finished clearing.

Ten years ago this region was a different machine. The cranes were shorter, the ships were smaller, and the slack that soaked up a four-day closure was still sitting inside the schedule, unpriced and largely unnoticed by the people booking boxes on it. That slack is gone. It went away with the bigger vessels, with the concentrated alliance calls and with the rebuilt terminals, all of which made the system far more productive per crane-hour and far less forgiving per bad week. That is the thing I want you to carry out of this, because it changes what a typhoon costs you, and it changes what you should do this week rather than next quarter.

Start with the plain facts, because everything you do downstream depends on getting the window right. Ningbo closed from 10 to 13 July. Shanghai closed across the same days, with prolonged terminal closures and vessel delays on top of the shutdown itself. Ten ports across China, Japan and Taiwan halted operations in that window. Six more across China and Taiwan ran at reduced capacity without closing outright. Everstream's report carries the date 23 July, which means the numbers inside it were taken ten days after the weather went through. Operations had gradually resumed by then. The queues had not.

A four-day stop at your own gateway is one problem. A four-day stop across ten ports is a different problem entirely, and the difference is the whole story. When one gateway shuts, the boxes that would have left it can often be quietly re-routed to a neighbour that stayed open, and the disruption becomes a trucking bill and a few days. When ten ports shut inside the same weather system, there is no neighbour to lean on. The cargo piles up in the same place it was always going to leave from, and when the gates open, every held box wants the same berth in the same week. Arrivals concentrate. The vessel that was going to carry your cargo is now carrying the cargo that should have sailed four days earlier, and yours is standing behind it.

The wait-time table in the 23 July report is the most useful object in the file, and it is worth reading as a picture rather than as a ranking. Shanghai sits at 3.2 days of average wait. Ningbo sits at 2 days. Busan and Yokohama are both at 1.5 days, with bunching flagged at each. Port Klang is at 1.4 days, with construction work scheduled from 23 July to 1 August affecting berthing arrangements, particularly for the largest vessels. Manila is at 4.7 days, the regional high, which is a useful reminder that the delta was not the only place in Asia carrying a queue in July. Ten days after the closure, the two Chinese gateways still showed the longest waits in North Asia, and that is the sentence I would underline if I were you.

Now the part that should make you sit up. Both ports closed for the same four days. Ten days later, Ningbo's queue is more than a day shorter than Shanghai's. I am not going to pretend the report tells me why, and there are several readings that fit. Shanghai's call density may simply be higher, so the same four-day stop leaves a longer line. More Ningbo cargo may have moved during the disruption than the published updates captured. The two may differ in yard capacity or in how quickly the terminals worked down the pile. What matters for you is practical rather than academic. If you are choosing between the two for an eastbound booking in the coming weeks, the published wait times are the only published evidence you have, and at this moment they point at Ningbo.

Put the map down for a moment and look at the decade instead, because the decade is where the real explanation sits. Ten years ago a cyclone of this size produced a bad week and a round of apologetic emails. The coast absorbed it because the system had room: more sailings, smaller ships, longer berth windows, and a schedule with genuine slack built into it that nobody had yet worked out how to sell. Since then the vessels have grown, the alliances have concentrated calls into fewer and larger arrivals, and the terminals have been rebuilt around automation and deep-water berths designed for ships that did not exist in 2016. Every one of those changes made the gateway more productive per crane-hour. Every one of them also removed a little of the room that used to swallow a disruption. The efficiency and the fragility came out of the same set of decisions, and nobody should be surprised that they arrived together.

Look at the geography once more, because it explains both why this coast carries so much and why it has so little room left. Ningbo sits on the seaward side of the bay with deep water at the quay and the manufacturing belt of the interior directly behind it, which is the oldest reason a port succeeds. Shanghai's containers run out through the island terminal at sea, reached by a single bridge, and through the river berths behind the city. Two gateways, one hinterland, and a decade of growth that pushed a very large share of China's eastbound volume through a stretch of coast that fits inside one weather system. Ten years ago the shares were different and the ships were smaller. What changed is not the storm. What changed is the network: fewer gateways carrying more boxes, on larger vessels, on tighter windows, with less room in the schedule for a bad week.

Here is what actually happens to your box, in plain sequence. It is gated in before the closure, or it is gated in during the closure and stands in the yard. The sailing it was booked on departs without it. Your box is rolled to the next sailing. If that sailing is itself oversubscribed because ten ports' worth of held cargo is chasing it, it rolls a second time. Each roll is roughly a week on transpacific eastbound, and two rolls is a fortnight. Kuehne+Nagel's update flagged post-typhoon vessel bunching across Chinese gateways and raised rollover risk on eastbound cargo, which is a professional way of describing exactly this sequence. There is nothing mysterious in it. It is a queue, and your position in the queue was set by the day you booked and by how much other cargo was standing in front of you.

Let me put numbers on it, and I will be explicit about which ones are mine. Take a shipper moving 40 FEU a month out of the delta on transpacific eastbound. Assume the month runs 30 days, so the booking flow is about 1.33 FEU a day. Four days of closure, 10 to 13 July, puts roughly 5.3 FEU, call it 5, into the pile that could not load on the day it was planned to load. Now assume, and this is my assumption rather than a published figure, that 25 percent of eastbound bookings out of the delta rolled at least once during the two weeks after reopening. On 40 FEU a month that is 10 boxes. Assume an average roll of 7 days. That is 70 container-days of delay arriving in your supply chain inside a single fortnight, from an event that lasted four days.

Turn those container-days into something you can feel. Assume your US distribution centre runs on 14 days of cover for the affected lines, and assume your own internal safety floor is 10 days. Ten boxes landing 7 days late is not a uniform 7-day hit across the whole range. It lands on the specific items those boxes were carrying. If one of those boxes was the replenishment for a line sitting at 14 days of cover, then at the moment you expected the stock to arrive you have about 7 days left, which is below your own floor. That is the moment you either pay to expedite or you stock out, and the decision is usually made in an afternoon by someone who did not plan for it. The trigger is not the typhoon. The trigger is the gap between 14 days of cover and a 10-day floor. A shipper holding 30 days of cover rides the identical event and notices nothing at all.

Another version of the same arithmetic, this time for the planners rather than the inventory people. Assume your door-to-door planning figure from the delta to a US west coast distribution centre is 35 days. Add the four-day closure and the 3.2-day average wait at Shanghai reported on 23 July, and you are at roughly 42 days. That is a 20 percent stretch on the assumption your whole calendar was built on. If your purchase orders, your production schedule and your customer promises were all set on 35, then all three are now wrong by about a week, and the fix is not a faster ship. The fix is re-dating all three before the next order goes in, which is cheap this week and expensive in six weeks when the shortage shows up on a shelf.

What you do in the days after a report like this is unglamorous, and most of it is about information rather than about cargo. Get the status of every open booking in writing. Not a portal screenshot that you interpret yourself, but an answer from the carrier or from your NVOCC that says which sailing the box is on and whether it appears on a rolled list. Ask for the sailing name and the departure date, not for a reassurance. Reassurances are free, and a named sailing is a commitment somebody can be held to. Then work out which of those boxes carry lines that are already thin, because those are the ones where a week matters, and a week on a line with 40 days of cover is not worth a phone call.

The next decision is Ningbo versus Shanghai, and you want to make it as a rule rather than as a scramble. For the next two months, quote and book every eastbound box out of your primary delta gateway against a named fallback sailing from the other gateway in the same week. That gives you something to move to when the first one rolls, and it also gives your counterparty the impression, correctly, that you have somewhere to go. The trap inside a dual-booking habit is not the second booking. It is the cancellation cut-off. Hold two bookings past the cancellation window and you pay for the one you do not use; cancel the wrong one and you have no booking at all. Decide 72 hours before cut-off, in writing, and give one named person the authority to make the call so that it does not become a discussion.

Alternatives deserve a paragraph, because people reach for them in the wrong order. Busan and Yokohama at 1.5 days are worth a look if you have cargo that tolerates transshipment, which means cargo with time in it and with no appointment waiting at the far end. Shifting to another Chinese gateway overland buys you a shorter queue and costs you a trucking leg plus a day or two of handover. Transshipment through Korea or Japan buys you a different queue and costs you a hub handling you do not control and a connection you did not choose. Air buys you time and costs a multiple of the ocean leg that nobody needs me to spell out. Each of those buys a different thing, so choose by what you are actually short of. Most people who reach for air in a week like this are short of inventory, not of speed, and the cheaper fix is usually the cover days.

The things that bite after a closure are rarely the closure itself. Free time and the demurrage clock start at gate-in or at discharge depending on how your contract words it, and a week of yard congestion converts into a bill nobody budgeted. Truck appointment windows at the terminal fill up before the berth does, so a box can be discharged and still not move for two days while the yard behind it is full. Equipment gets tight when every exporter in the region is pulling empties in the same fortnight, and a shortage of the right box type can delay a booking even after the berth queue has gone. Watch the gate and the equipment list as closely as you watch the berth, because the berth is the part that makes the news and the gate is the part that makes the invoice.

One more thing on the same theme, because it is where shippers lose the most money in events like this. Nobody tells you that you have been rolled. The booking stays open, the portal still shows a sailing, and the box quietly moves to the next voyage without an event that triggers anybody to call you. The only defence is a cadence: check the status of every open booking at a fixed time on a fixed day, twice a week while the backlog is clearing, and treat any booking whose sailing name has changed as a rolled box until proven otherwise. That is fifteen minutes of somebody's morning and it is worth more than any amount of escalation after the fact.

The six ports that were disrupted but not closed deserve a word, because partial disruption is easier to underestimate than a full stop. A port that never formally shuts does not appear on anyone's incident list, so nobody re-plans around it, and yet restricted operating hours, reduced gate windows and a yard running above its comfortable density will still push a booking out by days. In practice these are the ports where shippers get caught, precisely because there was no announcement to react to. If you have volume sitting behind any gateway in that second group, treat the four days as a soft closure rather than a nonevent, and ask your counterparty what the gate hours actually were on 11 and 12 July rather than assuming the published schedule applied.

There is a seasonal version of this as well, and it is the one that belongs in a planning document rather than in a crisis call. A cyclone that closes the delta for four days is not a rare event on this coast. It is a recurring feature of the calendar, which means the right response is not to treat July 2026 as an exception and move on. Write one four-day gateway closure into your annual plan as a baseline assumption, decide now how many cover days that implies for your top sellers, and price that decision once rather than paying for it badly every year. The map has not stopped producing these systems, and the network has less room for them each season, which is a trend you can plan against even though you cannot predict the individual storm.

Over the year rather than the week, the useful movement is concentration, and this event is a cheap way to measure yours. Write down what share of your eastbound volume sits behind a single Chinese gateway. If that number is above 60 percent, one four-day weather system can reach most of your book, and it did exactly that in July. Set a target to bring the top gateway below 60 percent by the time your 2027 contract round closes, and use the 23 July wait-time table as the reference you argue from: Ningbo at 2 days, Shanghai at 3.2, Busan and Yokohama at 1.5. Those are real published numbers, and a contract conversation grounded in them goes better than one grounded in general anxiety.

There is also a longer view, and it is the one I would take if I were sitting on your side of the desk. The delta will keep growing, the vessels will keep getting larger, and a four-day closure in July will keep costing more than the same closure cost in 2016, because the slack that used to absorb it has been spent on productivity that everybody wanted and nobody wants to give back. You cannot buy that slack back from the port. You can only hold it yourself, in cover days, in a second gateway that is actually used rather than merely listed, and in a booking rule that does not require anybody to make a judgement call at midnight. Ten years ago this coast was a different machine and a typhoon was a bad week. Today it is a network, and a bad week anywhere in it reaches you. The map changed. The part of the map you control can change with it.

↑ Back to the key points

— By Tomás Ferreira